Thread by @robgillezeau

We're back with day two of the #UnsettlingEconomics conference. A great lineup today: Krishna (SFU) & Ravi Pendakur (UO), Bryan Leonard (ASU), Jeffrey Burnette (Rochester), @MaggieECJones (UVic), Jasmin Thomas (FIN), @ProfPButton (Tulane) & Boyd Hunter (ANU)

External Tweet loading...
If nothing shows, it may have been deleted
by @robgillezeau view original on Twitter

Krishna and Ravi are kicking things off with their work, "How Do Modern Agreements and Opt-In Legislation Affect
Household Income Inequality in Aboriginal Communities?", looking at how agreements shift income / income inequality.

In particular, they're narrowing in on community level income inequality over time as influenced by signing of various agreements: Sectoral SGA, Standalone SGA, CLCA, fNLMA, FNMA, no agreement.

Confirms Aragon (2015)'s work finding that Comprehensive Land Claim Agreements boost income by ~$10,000 (or over 20%). No tangible income benefits for FNMA, FNLMA.

But what is this story missing? CLCAs may have a big income effect, but no real impact on income inequality. However, mixing CLCAs + SGAs, or opt ins (FNFMA FNLMA) cause large drops in income inequality.

What's going on in the back end? CLCAs decreases within-group inequality, but actually boosts inter-group inequality.

So the takeaways?
1) Modern agreements reduce inequality by as much as 5 percentage points off of Gini coefficients (a big change!)
2) CLCAs boost income, but not inequality
3) Opt ins and SGAs reduce inequality

Now we've got Bryan Leonard on his joint work w/ Dominic Parker and Terry Anderson on "Poverty from Incomplete Property Rights: Evidence from American Indian Reservations"

Focus on the Dawes Act and the attempt to privatize Native American property and its impact on Native American incomes.

Paper is pushing back on "traditional" narrative of: history -> land endowment -> income. Suggests that this story is a bit to simple.

Headline result: relationship between prime land and income is fundamentally different on reservations vs all US counties. Clear, upward sloping relationship off reserve while its a U shape on reserve.

Tying nicely into the "resource curse" literature with an alternative channel:
Prime land -> Dawes Act (1887) ->Institutions -> Incomplete Property Rights

Basic story: prime land on reservation was a target of privatization attempts, leading to incomplete property rights and thus lower income.

Initial set of regressions demonstrate this unqiue negative interaction of prime land on reservation causing income declines. Investigates the Dawes Act and its re-allotment of parcels of land as a mechanism.

Basic story seems to be: more prime land -> more likely to be allotted + done more quickly -> End of Dawes in 1934 + trust restrictions

Best places allotted first: mostly fee simple
Next best allotted later: stuck in trust
Worst never allotted: tribal control

Means reserves with good land end up substantially more fragmented with less autonomy, which drives the results.

Why? Checkerboarding creates overlapping governance regimes, prevents control of externalities, raises transaction costs of larger developments

So what are the poorest reservations? Those with no good land (no allotment) or all good land (rapid allotment) end up better off than reserves with mediocre land, which ends up heavily fragmented.

Clear policy implication: reestablish clear Indigenous tribal sovereignty over thee land.

Important takeaway: formal privatization with constraints can be a whole lot worse than informal property rights.

Next up: Jeffrey Burnette (joint work / Jason Younker and David Wick) on Statistical Termination or Fewer Self-Identified Students: What is causing the decline of American Indian and Alaska Native undergraduate enrollment?

Reviewing the history of US enumeration & race:
1860: first inclusion of American Indian as a category
1960: self indentificaiton of race begins with a limit of 1 identiyy
1980: 2 part question to reduce likelihood of Hispanic undercount
2000: multiple box self-identification

2007 guidelines from the Dept of Ed:
Hispanic: if Hispanic alone or in combination with an other racial category
American Indian or Alaska Native: if American Indian or Alaska Native alone and non-Hispanic
Two or more races: if more than 1 category selected + non-Hispanic

Pretty clear potential implications of these grouping categories on the salience of Indigenous identity in the academy

Raw data shows increasing enrollment + relatively flat share of % of enrollment until rules implementation re: racial definition in 2010. Substantial declines in both post-2010

Eyeballing the graphs it looks like as much as 25% of Indigenous enrollment disappears over this peirod

Burnette now digging into the micro data using the ACS and IPEDS: inferring single race identity from data. Finds that with this approach things are relatively stable over time.

So where are we actually seeing declines? It's among folks who are once-race bridged self-identifying / non-Hispanic. In other words, it seems likely that American Indians are increasingly identifying as multiple race.

Looks like my eyeballing was about right. Regressions find a 24.5% estimated undercount in Indigenous enrollment. Reported: 128k vs Adjusted Enrollment: 170k

Policy implication. The Department of Education has indicated they will not consider splitting out the multiple race categories so in as far as racial identity drives their policy direction there is now a systematic and substantial undercount of Indigenous students.

I'll take this opportunity to link to some relevant works. Here is Bryan's fabulous paper from earlier this morning on incomplete property right: drive.google.com/file/d/1_PiuT2…. This one is a must read imo.

Ravi and Krishna's paper isn't up online yet, but here's the first paper in this line of research on modern agreements that they're working on: sfu.ca/~pendakur/pape…

Jeffrey's paper isn't also online, but here's a neat paper he wrote providing some of the first evidence on the impact of the Great Recession on Native Americans that inspired some of my forthcoming work on the same topic: aeaweb.org/conference/201…

And, not to jump the gun, but @MaggieECJones is up next with her fabulous job market paper on student aid and the distribution of educational attainment: maggieecjones.files.wordpress.com/2017/12/jonesj…

And here we go with my soon-to-be colleague @MaggieECJones on her JMP.

Motivation / question of interest: how to rising costs of college impact educational attainment and labour market outcomes with a focus on Indigenous peoples.

How: identifies exogenous variation in costs through student aid

In particular, she takes advantage of reductions in targeted post-secondary support for Indigenous students in 1989 that i) limited per student funding and ii) decreased probability of receiving funding.

Uses variation in expected cost of higher ed across cohorts and ethnicities

Foreshadowing: this is going to have substantial impacts on the decision to finish high school.

Preview of findings:
i) $1k increase in cost -> 3pp decline in community college completion
ii) high graduation declines 1.7pp driven entirely y a 4.3pp decline on-reserve
iii) cutbacks explain ~10% of contemporary gap in hours worked between Indigenous & non Indigenous peoples

Back to the main story. The actual program she is discussing was launched in 1977 and was initially generous covering most costs.

Feds were worried that with changes in Bill C-31 (re discrimination against Indigenous women) costs would skyrocket so the program was heavily cut.

Most of will know the new, less generous program, the Post-Secondary Support Program, that was introduced in its place.

Maggie uses the 2006 Census of Population, combining date of birth, school attendance rules to see when people should have graduated. Look at eligible (FN/Inuit) vs non-eligible (Metis/non-Indigenous) for comparison.

Focuses on following schooling types: no school completed / HS / trade school / community college / bachelor's degree

Controls for tuition in province for each level of schooling at each time period.

Looking a the raw data: post program cutbacks, we see a bump in those not completing high school, increase in those only finishing high school + decline in all PSE programming completion

DID framework: controls for gender, tuition, tribe, CMA-province, birth quarter, latitude, longitude, distance to closet CMA, group FE, year FE

Big Impacts:
1.8% jump in NS
4.5% jump in HS
-3.0% decline in community college
-2.7% decline in university ed
-no statistically significant trades school impact

Why is the effect focused on reserve? Possibilities: low return to education on reserve, high u/e vs other areas, greater prevalence of low skill occupations

Putting the high school effects in context: in 1989 on-reserve graduate rates were 41.6% so a 4pp decline is quite substantial.

Labour market impacts over the longer run: 2.7pp decline in labour force participation + average decline of 1.028 hours worked per week.

Fascinating results on where this hours of work decline happens: largely among women in the middle quantiles of hours worked.

Takeaways on this paper:
-PSE grants targeted to Indigenous peoples really matter
-The prior decision to slash the program had large effects that are going to be persistent
-We can't ignore the impact of PSE supports on the decision to finish high school

Lunch break time. Live tweeting will be back in action in 50 minutes.

And we're back with Jasmin Thomas and her work on the impact of introducing formal childcare services on labour force participation in Inuit Nunangat.

Preview of results: jump in labour force participation for women, but not men. Impact greatest in Quebec. Positive impact on maintenance of Indigenous languages.

Unique feature that suggested the research topic: multifamily dwellings with many more children on average than non-Indigenous households

Programs being evaluated: FN and Inuit Child Care Initiative (FNICCI) and Aboriginal Head Start in Urban and Northern Communities (AHSUNC). FNICCI had 462 sites for 8500 children as of 2010, head start had 140 programs as of 2010.

Caveat: while this paper is a success story. Child care is still viewed as a chronic issue with ~700 people on wait lists in Iqaluit alone.

The treatment: Almost no communities have child care programs before 1995 and it surges over the 1995 - 2005 period. Pretty clean timing for evaluation purposes.

Employs a DID approach w/ standdard controls, language spoken, # of adults in household,, year and census subdivision FE

The female labour force participation bump looks substantial coming in at ~4pp.

Why is the effect concentrated in Quebec? Nunavik has $7 per day fees. In Nunavut, they range from $15 - $54 per day.

Robustness checks re: migration, household structure, and employment spillovers given small community size. No problems on any front.

Now onto language. Interesting preliminary results suggesting multiple years of childcare tenure boosts Indigenous language use at home. More work to do on this one.

I'm personally excited to see this data taken to Baker, Gruber, and @kevinmilligan results. Very curious about the Nunavik vs Nunavut divide and what shows up.

Next up is @ProfPButton from Tulane university on "Employment Discrimination against Indigenous Peoples in the United States: Evidence from a Field Experiment"

I've been looking for a resume study style set of results including Indigenous peoples and I'm thrilled that work is finally here.

Extremely limited evidence in this area so far. There is some ex-post survey evidence suggesting 31% of NA respondents believe they were discriminated against in hiring, but nothing of the empirical rigour we would usually expect.

Signal Indigenous status in a few ways: last name in some cases, first names in other, and then markers of Indigenous activities.

Oh, and I left out the language marker possibility or signal of being on a reservation. Patrick has done a great job and working out which signals may work where in the US and for which peoples.

What's the problem with names? Only 4 names can confidently be assigned to signal a particular background. (methodology here is in the paper)

What about names that "sound" Native American? They're actually incredibly rare in the data.

First names are also not particularly usable for the exercise, although there is enough commonality to use Native Hawaiians in some cases.

So how else is status suggested in resumes? Activities such as mentoring youth in the Native American community (i.e. Big Brothers, Big Sisters).

Last signal: language is a pretty clear market as non-Indigenous speakers are ~0

Occupations used: retail, servers, kitchen staff, janitors, security. Button has also done a thorough job at matching Indigenous signals to regions of sufficient population density.

70% no employer response. Of 30% that respond, which is split evenly between emails and calls. Any positive response is treated as a callback, but results are robust to alternative measures.

Regression controls for Indigenous status (NA, NH, AN), NA applicants x HS on reserve, HS in a rural area, other resume style and signal controls.

Results: white and Indigenous callback rates are not statistically different. If discrimination exists in the callback rate it would max out at ~1%. Surprising, but important result in the US context.

No differences by job, cities, signal types, use of multiple signals, etc

Is this a matter of signal salience? Unlikely, as use of similar signals in other studies for other minority groups picks up results. Explicitly tests salience as well using Mechanical Turk. Name signals are somewhat weak, especially nationally.

Signal only sort of works in Arizona and New Mexico. Oddly, people are viewing Navajo names as being foreign instead of NA.

Exciting next step: taking this study to Canada. I would be quite surprised if these results were the same north of the border. Plans on working with Indigenous groups the next time around to better inform the project. Looking forward to this project!

Last formal presentation of the day: Boyd Hunter from ANU on combining the profit motive and broader aspirations of Indigenous communities.

Review of Indigenous owned businesses in Australia: much smaller on average, 72.4% Indigenous workers (vs 0.74% in non-Indigenous businesses). Joint owned companies have a workforce that is still 46.9% Indigenous in the workforce.

Note that this sample is a procurement sample and not a general sample of Australian businesses.

Ending the day with a discussion of the today's papers and what the big research topics are going forward.

And that's a wrap! Back tomorrow morning with the concluding panel sessions.